This is an extremely bad idea. Your chances of getting some malware are probably more likely than not, after playing around with something like this for 10+ minutes...
How? The odds of hitting a site with a browser 0-day has to be extremely low, certainly not "more likely than not". Sure you might hit sites that try to get you to download malware, but just don't download anything.
Half the struggle in exploiting someone behind NAT/FW is getting them to engage with your infrastructure. Your attack surface is massively increased once you visit a website with your browser for instance.
I see other comments mentioning logging into random IPs over ssh. Now i trust the ssh client implementation more than most software, but it's easy to slip up and enable ssh agent forwarding for instance.